J. Am. Chem. SOC.1994,116, 11655-11666
11655
A Suite of Triple Resonance NMR Experiments for the Backbone Assignment of 1 5 N , 1 3 C , 2H Labeled Proteins with High Sensitivity Toshio Yamazaki,t**Weontae Lee,' Cheryl H. Arrowsmith," D. R. Muhandiram,? and Lewis E. Kay+?* Contribution from the Protein Engineering Centers of Excellence and Departments of Medical Genetics, Biochemistry, and Chemistry, University of Toronto, Toronto, Ontario, Canada M5S IA8, and Division of Molecular and Structural Biology, Ontario Cancer Institute and Department of Medical Biophysics, University of Toronto, 500 Sherboume Street, Toronto, Ontario, Canada M4X I K9 Received July 18, 1994@
Abstract: A suite of NMR pulse schemes is presented for the assignment of HN, lSN, 13Ca, and 13CBresonances in 15N,13C,2H labeled proteins. The experiments exploit the line narrowing of the 13C spins caused by the substitution of deuterons for bound protons and are therefore well suited for application to molecules with masses on the order of 30-40 kDa where standard triple resonance experiments may fail completely or provide spectra of poor sensitivity. The methods are demonstrated on a 37 kDa ternary complex of a 15N, 13C, and -70% 2H labeled sample of trprepressor, the corepressor 5-methyltryptophan, and a 20 basepair trp-operator DNA fragment. All of the experiments in the family presented provide well over 95% of the expected intra- andor inter-residue correlations. Carbon relaxation measurements of the complex indicate that the use of deuteration increases the average 13Ca Tz value from 16.5 to 130 ms.
Introduction The development of triple resonance multidimensional N M R spectroscopy has greatly increased the scope of proteins that can be studied by NMR The basic building block of a triple resonance experiment consists of a set of pulses and delays during which magnetization is transferred between heteronucleihomonuclei via the scalar coupling connecting the participating spin^.^^^ Because such couplings are considerably larger than homonuclear 'H-lH scalar couplings, the method can be applied to proteins which are significantly larger than would otherwise be possible using conventional 'H NMR techniques. Nevertheless, the application of some of the triple resonance experiments to proteins on the order of 40 kDa can still pose a significant challenge due in large part to short transverse relaxation times which limit the transfer efficiency between coupled spins. In the past years, a number of studies involving selective deuteration or perdeuteration of proteins in concert with 2D 'H'H spectroscopy or 15N-edited experiments have demonstrated the potential for sensitivity enhancementthrough the elimination of competing relaxation pathways that would otherwise be available to proton spins in fully protonated sample^.^-^ More recently it has been demonstrated that fractional substitution of Protein EngineeringCenters of Excellenceand Departments of Medical Genetics, Biochemistry, and Chemistry. Ontario Cancer Institute and Department of Medical Biophysics. Abstract published in Advance ACS Abstracts, December 1, 1994. (1) Bax, A.; Grzesiek, S. Acc. Chem. Res. 1993, 26, 131. (2) Clore, G. M.; Gronenbom, A. Science 1991, 252, 1391. (3) Ikura, M.; Kay, L. E.; Bax, A. Biochemistry 1990, 29, 4569. (4) Kay, L. E.; h a , M.; Tschudin, R.; Bax, A. J. Magn. Reson. 1990, 89, 496. (5) LeMaster, D. M.; Richards, F. M. Biochemistry 1988, 27, 142. (6) Shon, K. J.; Opella, S. J. Science 1991, 252, 1303. (7)Tsang, P.; Wright, P. E.; Rance, M. J. Am. Chem. SOC. 1990, 112, 8183. (8) Arrowsmith, C. H.; Pachter, R.;Altman,R. B.; Iyer, S. B.; Jardetzky, 0. Biochemistry 1990, 29, 6332.
*
@
deuterons for protons bound to 13C nuclei in combination with deuterium decoupling during periods where carbon magnetization evolves1° can significantly decrease the rate of carbon r e l a x a t i ~ n . ~ l - 'This ~ in turn results in an increase in the sensitivity and/or resolution of spectra that record carbon chemical shifts or in which carbon spins are involved in the relay of magnetization. Recent examples of the use of deuteration in concert with triple resonance spectroscopy include a 4D experiment which connects sequential amides in lSN, 13C, and 2H labeled proteins" as well as constant time 3D HNCA (CT-HNCA) experiment.13 In this paper we describe a suite of triple resonance experiments for backbone assignments in lSN, 13C, 2H labeled proteins. Included in this set of experiments are the CTHNCA,13CT-HN(CO)CA, HN(CA)CB, HN(COCA)CB, and 4D HNCACB. The first pair of experiments link HN, 15N, and 13Ca chemical shifts, while the second pair of experiments provide correlations connecting HN, lSN, and 13@ shifts. The 4D HNCACB experiment links the HN and lSNchemical shifts of a residue with intra- and inter-residue (13Ca, 13CB)pairs. The experiments are very similar to non-deuterium versions published p r e v i o ~ s l y ~but ! ~ Jare ~ designed to exploit the 13C line narrowing effect that deuteration provides. The methods are demonstrated on a 37 kDa ternary complex of a 15N, 13C,70% 2H labeled sample of trp-repressor (trpR), fully protonated 5-methyltryptophan, and a fully protonated 20 basepair trpoperator DNA fragment. The signal-to-noise (SM) of each experiment is evaluated, and the relative sensitivity differences (9) Reisman, J.; Hsu, V. L.; Parell, J.; Geiduschek, E. P.; Kearns, D. R. J. Am. Chem. SOC.1991, 113, 2787. (10) Browne, D. T.; Kenyon, G. L.; Packer, E. L.; Sternlicht, H.; Wilson, D. M. J . Am. Chem. SOC. 1973, 95, 1316. (11) Gresiek, S.; Anglister, J.; Ren, H.;Bax, A. J. Am. Chem. Soc. 1993, 115, 4369. (12) Kushlan, D. M.; LeMaster, D. M. J. Biomol. NMR 1993, 3, 701. (13) Yamazaki,T.; Lee, W.; Revington, M.; Mattiello, D. L.; Dahlquist, F. W.; Arrowsmith, C. H.;Kay, L. E. J. Am. Chem. Soc. 1994,116,6464. (14) Grzesiek, S.; Bax, A. J. Mugn. Reson. 1992, 96, 432.
0002-7863/94/1516-11655$04.50/00 1994 American Chemical Society
11656 J. Am. Chem. SOC., Vol. 116, No. 26, 1994 between recording spectra on a fully protonated vs a 70% deuterated sample are discussed. Finally, a comparison of the 13Ca Tz values for 13C spins one-bond coupled to either lH or 2H is presented.
Experimental Section Uniformly 15Nand 13Clabeled trpR was isolated from E. coli strain CY1507015containing the overproducing plasmid pJPR2 grown in 1-2 L of minimal media with I5NH4Cl and D-glUCOSe-'3Cs as the sole nitrogen and carbon sources, respectively. Purification was as described p r e v i ~ u s l y .For ~~~ triply ~ labeled 15N, 13C,and ZHtrpR the cells were adapted to growth in DzO as follows. Cells were first grown in M9 media with 200 p g / d ampicillin and 33% DzO as the solvent at 37 "C and then plated out onto an M9 agar plate made from 33% DzO. A colony from the 33% plate was used to inoculate M9 media with 56% D20, cultured overnight at 37 "C, then plated onto a 56% D20 M9 plate. For large-scale purification of triply labeled trpR, a single colony from the 56% plate was used to inoculate 10 mL, of 70% DzO M9 media containing 15NH4Cland ~-glucose-'~Cs.After ovemight growth this culture was used to inoculate 1.5 L of the same M9 media. Although the cells grew a factor of 2 times more slowly than on normal M9 media, induction and purification yielded 60 mg of pure 15N, I3C, ZHprotein, similar to the amount normally obtained from H20 growths. Protein-DNA complexes were prepared by adding the appropriate amount of synthetic operator DNA16 to trpR followed by dialysis and concentration into a pH 6 solution of 50 mM sodium phosphate and a 4.8 mM racemic mixture of 5-methyltryptophan. Only the L-isomer has significant affinity for trpR.17 Final NMR sample conditions were 2.4 mM trpR (trpR is a symmetric dimer of 108 amino acid monomers; 2.4 mM refers to the subunit concentration), 1.2 mM palindromic double stranded operator DNA, 2.4 mM 5-methyl-~-tryptophan,and 50 mM sodium phosphate adjusted to pH 6. The 15N,I3C fully protonated Q R sample used in the NMR studies was a factor of 2.6 lower in concentration relative to the ISN, 13C,ZHsample. All experiments were performed on a Varian UNITY 500 MHz spectrometer equipped with a pulsed field gradient unit and an actively shielded triple resonance probe head. Four separate synthesizers were used, one for the generation of each of 'H, 13C, 15N, and 2H pulses. The gating and modulation of rf from the 15Nand zHsynthesizers were controlled by a single channel, requiring a number of small modifications to our spectrometer. On our spectrometer it was not possible to disable the 2H lock receiver during either 2H decoupling or the application of field gradient pulses; however, for improved spectrometer stability, the lock receiver should be gated off during ZHdecoupling or gradient pulses if possible. For the constant time HN(CO)CA(CTHN(C0)CA) experiment, a matrix of 84* x 28* x 512* points (the * refers to complex points) was acquired with acquisition times of 25.7, 23.0, and 64.0 ms in each of tl, t2, and 13. The corresponding spectral widths in FI,Fz, and F3 were 3268, 1217 and 8000 Hz. The total measuring time using eight scans per FID and a repetition delay of 2 s was 45 h. The HN(C0CA)CB experiment was acquired as a 64* x 28* x 512* complex matrix with spectral widths of 9000, 1217, and 8000 Hz in F1, Fz,and F3 corresponding to acquisition times of 7.1, 23.0, and 64.0 ms in each of t1, tz, and t 3 . A total of 16 scans were acquired per FID with a repetition delay of 1.8 s to give a total acquisition time of 62 h. The HN(CA)CB experiment was acquired as a 64* x 28* x 512* complex matrix with acquisition times of 8.3, 23.0, and 64.0 ms in each of t l , 12, and t3. Spectral widths of 7650, 1217 and 8000 Hz in F1, Fz,and F3 were employed. A total of 16 scans were acquired per FID with a repetition delay of 1.8 s to give a total acquisition time of 62 h. The 4D HNCACB experiment was acquired as a 16* x 32* x 28* x 512* complex matrix with acquisition times of 7.1, 5.5, 23.0, and 64.0 ms in each of t l , tz, t 3 , and t4 and spectral widths of 2262, 5780, 1217, and 8000 Hz in F1, Fz,F3, and F4. Two scans were acquired per FID with a repetition delay of 1.8 s to give a total acquisition time of 124 h. The CT-HNCA data set was acquired as described p r e v i o ~ s l y . ~ ~ (15) Paluh, J. L.; Yanofsky, C. Nucleic Acids Res. 1986,14, 7851. (16)Lefevre, J. F.; Lane, A. N.; Jardetzky, 0. Biochemistry 1987,26, 5076. (17)Marmostein, R. Q.;Joachimiak, A.; Sprinzl, M.; Sigler, P. B. J . Bioi. Chem. 1987,262, 4922.
Yamazaki et al. All spectra were processed using nmrPipe/nmrDraw software developed by F. Delaglio, N I H , I 8 and analyzed using the program PlPP.19 In the acquisition dimension a solvent suppression filter was applied to the data to minimize the water signalz0prior to apodization with a 65O-shifted squared sine-bell window function. The data were subsequently zero filled to twice the size and Fourier transformed. Only the downfield half of the spectrum was retained for the 3D data sets. In order to minimize space and processing time in the case of the 4D data set, only the HN region of the spectrum was retained in F4 (7-10 ppm). In the case of the HN(C0)CA data set, the data were apodized in the I3Ca dimension using a 65O-shifted squared sine-bell window function, zero filled to 256 complex points followed by Fourier transformation, phasing, and elimination of the imaginary half of the signal. The size of the I5N time domain was doubled via mirror image linear prediction?[ apodized using a cosine-bell function, zero filled to 128 complex points, Fourier transformed, and phased, and imaginaries were eliminated. The absorptive part of the 3D data set was 256 x 128 x 512 real points. In principle, the 13Catime domain could be extended via linear prediction as well, although this was not done in the present case. For the HN(C0CA)CB and HN(CA)CB data sets, the data were apodized in the carbon dimension using a 65"-shifted squared sine-bell window function and zero filled to 128 complex points followed by Fourier transformation, phasing, and elimination of the imaginaries. In the case of the HN(C0CA)CB spectrum a zero-order baseline correction was applied to the carbon dimension (after transformation and phasing) to correct for the dc offset introduced by the fact that the first point was not acquired at t = 0. The I5Ndimension was processed as described above to give a final absorptive data set of 128 x 128 x 512 points. For the 4D HNCACB spectrum the data were apodized in the I3CB dimension using a cosine-bell window function, zero filled to 64 complex data points, Fourier transformed, and phased and imaginaries discarded. In the 15N dimension, after application of a cosine-bell window function, the data were zero filled to 64 complex data points, Fourier transformed, and phased and the imaginaries eliminated. Finally, the 13Catime domain was extended to 32 points by mirror image linear prediction, apodized using a cosinebell window function, zero filled to 64 complex data points, Fourier transformed, and phased and the imaginaries were discarded. The center of the I3Ca spectrum was shifted from 45 to 59 ppm by a frequency domain circular shift.18 The final absorptive part of the data set consisted of 64 x 64 x 64 x 231 real points. HN TIrelaxation experiments were performed on the fully protonated sample using delays (7') of 0.1, 0.3, 0.5, 0.7, 1.5, 2.0, and 4.0 s. For the deuterated sample, delays of 0.2, 0.6, 1.0, 1.4, 2.0, 3.0, 4.0, and 8.0 s were employed. The TlPmeasurements were performed with the carbon carrier at 58 ppm using a 1.4 kHz spin lock field. Spin lock times of 4, 8, 12, 16, and 20 ms were used for the protonated sample, while times of 16,32,48,80, 120, and 160 ms were employed for the deuterated sample.
Results and Discussion The principal motivation for the substitution of zH for carbonbound 'H spins in 15N and 13C labeled proteins is the decrease in 13C relaxation efficiency due to the approximately 7-fold lower gyromagnetic ratio of *Hcompared to 1H.10-13Figure 1 shows the expected contributions to the carbon line width from both dipolar and chemical shift anisotropy (CSA) relaxation mechanisms for an isolated 13Ca-lHa or 13Ca-2Ha spin pair as a function of isotropic molecular correlation time. Clearly the decrease in carbon line width as a result of the substitution of ZHfor 'H is significant. For molecules with correlation times, z, satisfying the relation z, 2 10 ns, the carbon line width is decreased by a factor of -15 for a 13Caone-bond coupled to a deuteron relative to a proton. The dramatic decrease in carbon relaxation rates with deuteration and the realization that a set (18) Delaglio, F. nmrPipe Sysrem of Sofhuare; National Instiktes of Health: Bethesda, MD, 1993. (19) Garrett, D. S.; Powers, R.; Gronenbom, A. M.; Clore, G. M. J. Magn. Reson. 1991,95, 214. (20) Marion, D.; h a , M.; Bax, A. J . Magn. Reson. 1989,84, 425. (21) Zhu, G.; Bax, A. J . Magn. Reson. 1990,90, 405.
Backbone Assignment of 15N, I3C, and 2H Labeled Proteins 301
1
I
J. Am. Chem. SOC., Vol. 116, No. 26, 1994 11657 to-noise (S/N).28,29In addition, saturation/dephasing of water magnetization is kept to a minimum by ensuring that (i) homospoil gradient pulses are applied only when the water magnetization is stored along the z axis and that (ii) the water magnetization is placed along the +z axis immediately prior to detection. In this context, each pulse scheme makes use of selective water pulses as described p r e v i o ~ s l y . ~ ~ - ~ ~ The flow of magnetization transfer in each of the experiments can be described concisely by
HN % MN
13 a
C (rl)
&Cn
-
15N (t2)
JHN
HN 0 3 ) (3D CT-HNCA)
b C
10 12 14 Correlation time (ns)
16
18
20
Figure 1. Expected contributions to the I3Ca line width from both dipolar and CSA relaxation assuming an isolated I3Ca-lHa or I3Ca2Ha spin pair attached to a molecule tumbling isotropically in solution. Curve a is the contribution to the line width from dipolar interactions with a directly attached 'H at a distance of 1.10 A;42curve b illustrates the dipolar contributions from a directly attached deuteron; curve c shows the contributions from CSA assuming an axially symmetric chemical shift tensor with ha = 34 ppm.4l
of high-sensitivity experiments can be derived which exploits this result has led us to develop the suite of pulse schemes illustrated in Figure 2a-d for the backbone assignment of proteins that are 15N, 13C, 2H labeled. An additional member of this family of experiments,the CT-HNCA, has been published p r e v i ~ u s l y .Because ~~ all of the experiments make use of the "out and back" approach for magnetization transfer whereby magnetization both originates and is detected on the HN spin and since none of the transfer steps or evolution periods involve aliphatic/aromatic lH magnetization, these experiments will work most efficiency for samples with significant levels of deuterium enrichment. The temptation to prepare a fully deuterated sample is best resisted however since complete deuteration of a sample precludes its use for experiments which require that magnetization originate or be detected on carbonbearing protons. In addition, full deuteration will likely result in a significant increase in HN TI relaxation times which would necessitate the use of longer relaxation delays. For this reason, it is unlikely that a fully deuterated sample will prove optimal for structural studies. We have therefore produced a 70% 2H labeled trpR sample to be used in the present studies, and where necessary (see below), the experiments have been optimized for samples that are not fully labeled in deuterium. Because the sequences are closely related to 15N, 13C triple resonance experiments that have been discussed in detail in the literature previously, we will not attempt to provide a quantitative explanation of each experiment here. Rather we highlight the unique features of these pulse sequences and provide a rationale for their design. Pulsed field gradients are employed in all of the experiments in order to minimize artifacts and the residual water in ~ p e c t r a ~as~ -well ~ ~ as to select for the coherence transfer pathway whereby magnetization passes through nitrogen in the final portion of each of the sequence^.^-^^ An enhanced sensitivity approach is used to optimize signal(22) Hurd, R. E.; John, B. K.J. Magn. Reson. 1991, 91, 648. (23) Von Kienlin, M.; Moonen, C. T. W.; Van Der Toom, A.; Van Zijl, P.C. M. J . Magn. Reson. 1991,93, 423. (24) Bax, A.; Pochapsky, S. J . Magn. Reson. 1992,99, 638. (25) Kay, L. E.; Keiffer, P.;Saarien, T.J . Am. Chem. SOC.1992,114, 10663. (26) Schleucher, J.; Sattler, M.; Griesinger, C. Angew. Chem., Int. Ed. Engl. 1993,10, 32. (27) Muhmdiram, D . R.; Kay, L. E. J . Magn. Reson. B 1994,103,203.
13
CO(j-1,
-
-
15
JNC'
N (4)
JHN
03)
(3D CT-HN(C0)CA)
-
J'w 15N J".13co(i-l)
JCTa
13ca (i-1)
JHN
'N (t2)
JHN
HN (f3)
-HN
'N ( t 3 )
Jcc
(3D HN(C0CA)CB)
HN (t3)
I5N (t2)
-
JHN
(f4)
(3D HN(CA)CB)
(4D HNCACB)
For completeness, the transfer path in the CT-HNCA experiment13 is provided as well. The relevant couplings involved in each of the magnetization transfer steps are indicated above the arrows, and tl, t2, r3, and t4 denote acquisition times. The CT-HNCA and CT-HN(C0)CA experiments provide correlations of the form (I3Ca, 15N,HN) with the 13Ca and 15N chemical shifts recorded in a constant time Note that the HNCA experiment provides both intra- and inter-residue correlations, while the HN(C0)CA provides only sequential connectivities. In both of the CT-HNCA and CT-HN(C0)CA experiments, 13Cmagnetizationis recorded during a delay which is set to l/(JCc), where JCC is the one-bond aliphatic carboncarbon scalar coupling. The increased acquisition time in the carbon dimension (-27 ms) relative to previous versions of these e ~ p e r i m e n t s ~(-8 * ~ Jms) ~ coupled with the fact that the significant one-bond homonuclear 13Ca- 13CBcoupling does not modulate the time domain carbon signal in the constant time experiments results in over three times the resolution in this dimension. The choice of a constant time delay of l/(JCc) 27 ms ensures that the majority of correlations in spectra of
-
(28) Cavanagh, J.; Palmer, A. G.; Wright, P. E.; Rance, M. J . Magn. Reson. 1991.91. 429. (29)Palmer, A. G.; Cavanagh, J.; Wright, P. E.; Rance, M . J . Magn. Reson. 1991, 93, 151. (30) Grzesiek, S . : Bax, A. J . Am. Chem. SOC.1993,115. 12593. (31) Stonehouse, J.; Shaw, G . L.; Keeler, J.; Law, E. J . Magn. Reson. A 1994,107, 178. (32) Kay, L. E.; Xu, G. Y.; Yamazaki, T. J . Magn. Reson. A 1994,109, 129. (33) Santoro, J.; King, G. C. J . Magn. Reson. 1992,97, 202. (34)Vuister, G. W.; Bax, A. J . Magn. Reson. 1993,101, 201.
11658 J. Am. Chem. SOC.,Vol. 116, No. 26, 1994
Yamazaki et al.
Figure 2. Pulse schemes of the (a) 3D CT-HN(CO)CA, (b) 3D HN(COCA)CB, (c) 3D HN(CA)CB, and (d) 4D HNCACB experiments. AU nmow (wide) pulses are applied with a flip angle of 90' (180'). Unless indicated otherwise, pulses are applied along the x axis. The 'Hand 15N carriers are centered at 4.7 ppm (water) and 119 ppm, respectively. Proton pulses are applied using a 23 lcHz field with the exception of the water selective 90" pulse during the fmt INEPT transfet" (indicated by the shaped 'Hpulse in the sequences) which is applied either as a 2 ms rectangular pulse or as a 2 ms pulse with the SEDUCE-14 profile, the 6.8 kHz f y pulses flanking the WALTZ-16x45decoupling periods and the 6.8 kHz WALTZ-16,45 decoupling sequence. (Note that in WALTZ-lQ all of the rf pulses are of phase x or -x.) The 15Npulses are applied using a 5.2 kHz field, with 15N decoupling during acquisition using a 1 kHz field. AU carbon pulses were applied using only a single frequency source. All decoupling is interrupted during the application of the gradient pulses.& In the case of 'H decoupling, the water magnetization must first be "returned' to the z axis prior to application of the gradient pulse and subsequently rotated to the decoupling axis prior to the continuation of
Backbone Assignment of 15N, 13C, and 2HLabeled Proteins
J. Am. Chem. SOC.,Vol. 116, No. 26, 1994 11659
Figure 2 caption continued deco~pling.~~ Sequence a, CT-HN(C0)CA: Carbonyl 180" pulses are applied using the SEDUCE-lez profde with a peak amplitude of 4.72 kHz corresponding to a length of 230 ps. The pulses are centered at 176 ppm. At point a in the sequence the carbon carrier is jumped from 176 to 58 ppm and subsequently retumed to 176 ppm at point b. Therefore the I3C' 180" pulses between points a and b are applied as phase modulated p ~ l s e s . ~The ~ . 13C' ~ ~ and 13Ca90" pulses are applied using a field strength of 3.93 kHz so that application of I3C' pulses minimizes excitation of I3Ca spins and vice versa.4 All I3Ca 180" pulses (8.78 kHz) with the exception of the 180" pulse during the 2Tc period (400 ps R E - B W 9 pulse with a peak amplitude of 15.7 kHz, affecting both I3Ca and I3CP spins) are applied as phase-modulated rectangular pulses since the carbon carrier is centered at 176 ppm during application of these pulses. The positions of the Bloch-Siegert compensation pulsesMare indicated by the arrowheads in the figure. I3Ca decoupling is achieved using a 118 ppm cosine-modulated WALTZ1645field employing pulses having the SEDUCE-150profile (330 ps 90" pulses). *Hdecoupling is achieved using a 0.9 lcHz WALTZ-1645sequence centered at 4.5 ppm. The delays employed are ta= 2.3 ms, t b = 5.5 ms, tc= 12 ms, t d = 1.7 ms, re = 4.1 ms, Tc = 13.8 ms, TN = 12 ms, 6 = 0.5 ms. The phase cycling employed is 41= (x, -x); & = 2 (x),2(-x); & = x ; 4 4 = 4(x),4(-x); 45 = x; rec = (x,-x,-xj). Quadrature in F1 is achieved via States-TPPIS1of &. For each value of t2, N- and P-type coherences are obtained by recording two data sets whereby the sign of the gradient gll is inverted and 180" added to the phase of 45 for the second data set. Data sets obtained in an interleaved manner for positive and negative gll values are stored in separate memory locations and pure absorptive line shapes in the 15Ndimension generated by adding and subtracting the N- and P-type data sets, storing the new data sets separately and applying a 90" zero-order phase correction in the acquisition dimension to one (not both) of the new data sets.25 The phase of 43 and the phase of the receiver are incremented by 180" for each increment of The durations and strengths of the gradients (rectangular) are gl = (0.5 ms, 8 G/cm), g2 = (0.5 ms, 4 G/cm), g3 = (1.0 ms, 10 G/cm), g4 = (1.0 ms, 1 G/cm), g5 = (1.0 ms, 7 G/cm), g.5 = (1 ms, -15 G/cm), g7 = (0.2 ms, 20 G/cm), gs = (1 ms, 8 G/cm), g9 = (1 ms, 0.4 G/cm), glo = (0.6 ms, 10 G/cm), gll = (1.25 ms, 30 G/cm), gll' = (0.125 ms, 27.8 G/cm), gl2 = (0.5 ms, 8 G/cm), g13 = (0.3 ms, 2 G/cm). Note that the effect of all of the IH pulses on water is such that the water magnetization is restored to the +z axis at the start of acquisition. Sequence b, HN(C0CA)CB: Many of the details of this sequence are similar to those of sequence a. Only the differences are described here. At point a in the sequence the carbon carrier is jumped from 176 to 43 ppm and subsequently retumed to 176 ppm at point b. The 13C"@pulses marked a (8.78 W z field) are applied as phase-modulated pulses so that the center of the excitation is at 58 ppm. Additionally, 13Cadecoupling is achieved using a 118 ppm cosine-modulated WALTZ-1645field employing pulses having the SEDUCE-lS0profile as before (i.e., the center of the decoupling field is at 58 and not 43 ppm). AU other 13C@pulses are applied so that the excitation is centered at 43 ppm. The l3CdP 90" pulses are applied using a 4.25 kHz field, while the 180" pulses are applied with a 9.51 kHz field to minimize excitation of carbonyl magnetization? Since I3CP (not ITa) chemical shifts of all the amino acids with the exception of glycine are recorded during tl, it is not necessary to apply an 15N 180" pulse in the center of the tl evolution period. If a smaller value of zfis employed to obtain both 13Caand I3CP shifts or for (very) slightly better sensitivity of glycine cross peaks, a 180' 15Npulse can be applied in the middle of the t l period for those values of t1 satisfying the relation tl > t180, where t 1 is~the I5N 180" pulse width. For t l < ~ 1 ~the 0 , I5N 180" pulse is applied immediately after the 13C@pulse of phase 4 4 . The application of the 15N 180" pulse in this manner minimizes baseline distortions in the carbon dimension caused by a nonideal initial tl value. It is possible to insert a C' 180" pulse in the center of the tl period to improve (slightly) the sensitivity of correlations involving glycine or asparagine/aspartic acid residues; however, because of the length of such a pulse, it is necessary to back-predict the first couple of points in this dimension. The delay tr is set to 6.8 ms. The phase cycling employed is 41 = (x,-x); & = 2(x),2(-x); (p3 = 4(y),4(-y); 44 = 8(y),8(-y); 45 = x; 46 = 4(x),4(-x); & = x; rec = (x,-x,-XJ). Quadrature in F1 is achieved via States-TPPISLof & and &. For each value of tz, N- and P-type coherences are obtained by recording two data sets whereby the sign of the gradient g12 is inverted and 180" added to the phase +, for the second data ~ e t . 2In ~ addition the phase 45 and the phase of the receiver are incremented by 180" for each new value of t2.51The durations and strengths of the gradients (rectangular) are gl = (0.5 ms, 8 G/cm), g2 = (0.5 ms, 4 G/cm), g3 = (1.0 ms, 10 G/cm), g4 = (1.0 ms, 1 G/cm), g5 = (1.0 ms, 7 G/cm), g6 = (1 ms, 15 G/cm), g7 = (1 ms, 1 G/cm), gs = (0.5 ms, 10 G/cm), g9 = (1 ms, 8 G/cm), glo = (1 ms, 0.4 Gkm), gll = (0.6 ms, 10 G/cm), g12 = (1.25 ms, 30 G/cm), g12' = (0.125 ms, 27.8 G/cm), g13 = (0.5 ms, 8 G/cm), g14 = (0.3 ms, 2 G/cm). Sequence c, HN(CA)CB: All 13C@pulses are applied with the "C carrier centered at 43 ppm using an 18 kHz field. I3C' decoupling is achieved using a 133 ppm cosine-modulated WALTZ-1645field employing pulses having the SEDUCE-150 profile (i.e., the carbon carrier is at 43 ppm). The delays used are as given above. For (very) slightly better sensitivity for glycine residues or if a shorter tfvalue is employed, a 180" I5N pulse can be inserted in the center of the tl period as described for the HN(C0CA)CB experiment. The phase cycling employed is 41 = (x,-x); & = 2(x),2(-x); & = 4(y),4(-y); 4 4 = 8(y),8(-y); 45 = x ; 46 = 4(x),4(-x); +, = x; rec = (x,-x,-x,x). Quadrature in FI is achieved via States-TPPIslof & and &. Quadrature in Fz is achieved using the gradient approach described above with the phase & and the sign of the gradient gs inverted to generate N- and P-type data setsz The phase 45 and the phase of the receiver are incremented by 180" for each new value of t2.51 The durations and strengths of the gradients (rectangular) are gl = (0.5 ms, 8 G/cm), gz = (0.5 ms, 4 G/cm), g3 = (1.0 ms, 10 G/cm), g4 = (0.5 ms, 8 Gkm), g5 = (0.1 ms, 20 G/cm), g6 = (0.1 ms, 30 G/cm), g7 = (0.6 ms, 10 G/cm), g8 = (1.25 ms, 30 G/cm), gg = (0.125 ms, 27.8 Gkm), g9 = (0.5 ms, 8 G/cm), glo = (0.3 ms, 2 G/cm). Sequence d, 4D HNCACB: Almost all details are the same as those for HN(CA)CB. The carbon carrier is centered at 45 ppm and 13C' decoupling achieved using a 131 ppm cosine-modulated WALTZ-16" field employing pulses having the SEDUCE-150profile. The value of TC is set to 6.8 ms. The phase cycling employed is 41 = (x,-x); & = y; 43 = x ; 4 4 = x; 45 = (x,-x); 4 6 = x ; rec = (x,-x). Quadrature in F1 is achieved via States-TPPP of 41. Quadrature in FZis achieved via States-TPPI of 4 2 and &. Quadrature in F3 is achieved using the gradient approach described above with the phase 4 6 and the sign of the gradient g8 inverted to generate N- and P-type data sets.25 The phase 4 4 and the phase of the receiver are incremented by 180" for each new value of hS1The durations and strengths of the gradients are as described for the HN(CA)CB sequence. proteins that are fractionally deuterated arise from 13C spins that are one-bond coupled to deuterons and not protons. Assuming isotropic molecular tumbling with a correlation time of 15 ns (MW -30 000) and including only the contribution to the relaxation of the 13Ca spin from dipolar interactions with the directly attached 'H or *H spin, 13Ca Tz values of 16 ms (13C-lH) and 249 ms (13C-2H) are calculated. Thus, during a constant time period of l/(JCc), the intensities of cross peaks arising from 13Ca-2H as opposed to 13Ca-lH pairs are roughly in the ratio of 5 to 1, assuming 50% fractional deuteration at the Ca position. For those residues displaying increased mobility and subsequently increased 13Ca T2 values, it is very likely that cross peaks of significant intensity will arise from 13Ca spins coupled to either 'H or 2H. Because of the significant one-bond deuterium isotope shift of approximately 0.4 ppm,' 1-13
the doubling of signals degrades spectral resolution in the carbon dimension. In order to avoid this problem in the CT-HN(C0)CA scheme of Figure 2a, we have made use of an approach described p r e v i ~ u s l y 'whereby ~ carbon magnetization arising from 13Ca-lHa pairs evolves during the constant time period 2Tc for a duration 2 r d = 1/(wHC). In this way this magnetization is antiphase immediately after the 2Tc period and is not refocused into an observable signal by the remaining pulses. Figure 2, parts b and c, illustrate the HN(C0CA)CB and HN(CA)CB pulse sequences, respectively. The HNCACB experiment was originally developed by Wittekind and MuelleS5 based on the CBCANH scheme of Grzesiek and Bax36 and provides correlations of the form (13Ca/i3@, 15N, HN). As in the HNCA experiment, correlations of both the intra- and the inter-residue variety can be obtained from this experiment. In contrast the
11660 J. Am. Chem. SOC., Vol. 116, No. 26, 1994 HN(C0CA)CB experiment provides correlations linking sequential shifts exclusively. Note that cross peaks linking either the 13Ca or 13CB shifts are proportional in intensity to cos2(2zJcczf) or sin2(2zJcczf), respectively. Therefore by tuning the value of zf suitably it is possible to select for either 13Caor 13@ cross peaks. For 15N, 13Clabeled samples, Wittekind and M ~ e l l e rand ~ ~Grzesiek and Bax36have chosen zf 1/(8Jcc) so that both sets of peaks are obtained, albeit at half of the maximum intensity possible neglecting relaxation losses and pulse imperfections. Because of the rapid 13Ca relaxation that occurs in fully protonated samples, it is not practical to choose tf 1/(4Jcc) to optimize transfer to the l3CB carbon in any event. In contrast, for studies with the 15N, 13C,70% 2H sample of trpR described here, we prefer to obtain (13Ca, 15N, HN) correlations from the CT-HNCNCT-HN(C0)CA experiments described above and (13CP, 15N, HN) correlations from HN(CA)CB/HN(COCA)CB experiments optimized for transfer of magnetization from 13CUto 13@ {zf 1/(4Jcc)}. For this reason we have referred to these experiments as HN(CA)CB and HN(C0CA)CB in keeping with the nomenclature introduced by Ikura et al.3 Neglecting relaxation and pulse imperfections, the sensitivities of the (13CP, 15N, HN) peaks in the HNCACB experiment with zf 1/(8Jcc) are reduced by a factor of 2 relative to cross peaks recorded in the HN(CA)CB experiment. In a similar vein, (13Ca, 15N, HN) cross peaks in the HNCACB experiment recorded with zf 1/(8Jcc) are a factor of 2 reduced in intensity relative to the corresponding peaks obtained from an HNCA experiment recorded in exactly the same manner. Thus for equal S/N the HNCACB experiment which provides both 13CUand 13CBcorrelations {zf 1/(8Jcc)}must be recorded for a period which is twice as long as the combined measuring times of the HNCA and HN(CA)CB {tf 1/(4Jcc)} experiments. Similar conclusions hold for the HN(C0)CACB experiment {zf 1/(8Jcc)} vs the HN(C0CA)CB {zf 1/(4Jcc)} and HN(C0)CA experiments. As will be illustrated later, for proteins with significant levels of deuterium enrichment (such as the 70% 2H sample considered in the present study), higher quality spectra are obtained by recording pairs of experiments which optimize a specific magnetization transfer rather than recording a single spectrum where the sensitivity of all correlations is compromised. As has been described above, the HNCA and HN(C0)CA experiments are recorded with constant time in the carbon dimension. This is particularly important given the (often) poor dispersion in the 13Cadimension of protein spectra. In contrast, we have found that for the level of enrichment of 2H in the trpR sample that we are studying (-70%) experiments which measure the 13CP chemical shift using a similar constant time strategy suffer significantly from sensitivity losses relative to their nonconstant time counterparts. Note that, in the HN(CA)CB/HN(COCA)CB experiments, cross peaks providing the 13@ chemical shift arise from magnetization that is transferred from l3Ca to 13CP. The process of completely transferring magnetization from 13Ca to 13CB and then back (after 13CP tl evolution) dictates that the magnetization resides on the 13CUspin in the transverseplane for -27 ms. In the case of constant time carbon evolution, magnetization residues further on the /?-carbon for -27 ms (l/JCc). For high sensitivity it is crucial that all carbons involved in the magnetization transfer process (in this case 13Ca and 13CB)be fully deuterated. In the case of -70% deuteration, only -35% of the population of molecules satisfy this constraint, assuming random deuteration, and the subsequent degradation of sensitivity unfortunately precludes the use of constant time carbon evolution. Since a large fraction of the signal comes from 13CB spins that are coupled to protons in this case, it is
Yamuzaki et al. 20
CT-HNCA
+-
-
16 0
-
13
C
Ii
I
-
(PPm) 0 0
40
-
-
-
-
-
-
(35) Wittekind, M.; Mueller, L. J . Magn. Reson. B 1993, 101, 201. (36) Grzesiek, S.; Bax, A. J . Magn. Reson. 1992,99, 201.
1 RZI
FZZ
V23
D24
LU
L26
K27
NZB
A29
Y30
Q3I
N32
D33
L34
H35
Residue
Figure 3. Strip plot showing the region of the CT-HNCA spectrum of trpR complex from R21 to H35. Both intra- and inter-residue correlations are available from this experiment. The sequential
connectivity derived from the analysis of the experimentspresented is indicated here and in Figures 4-6 by lines connecting the appropriate intra-/inter-residuecross peaks. necessary to employ 'H decoupling during the recording of carbon magnetization in the HN(CA)CB or HN(C0CA)CB experiments. Recall that in the case of experiments which record only 13Ca chemical shift in a constant time manner 'H decoupling is not necessary during the constant time evolution since the majority of the detected signal originates from carbons that are one-bond coupled to *H and not 'H. Moreover, a purging scheme is employed which selects for the 13Ca-2Ha population of molecule^.^^ Clearly such a purging scheme should not be employed in the case of the HN(CA)CB or HN(C0CA)CB experiments. In any event, the significantly lower resolution in the carbon dimension in the 13CPexperiments due to the coarse digitization used obviates the need to employ a purging scheme to eliminate the extra peaks caused by the deuterium isotope shift. Figures 3-6 demonstrate the applications of the various 3D pulse schemes to the ternary complex of trpR, 5-methyltryptophan, and a 20 basepair trp-operator DNA sequence (MW 37 000). In particular, strip plots comprising residues R21H35 are illustrated, taken at the 15Nand HN frequencies of the residues indicated. Clearly the sensitivity afforded by these experiments is excellent. In the figures illustrating the HN(C0CA)CB and HN(CA)CB experiments (Figures 5 and 6, respectively), a portion of the 13Ca chemical shift region has been included to indicate that it is possible to achieve essentially complete transfer of magnetization from 13Ca to 13CP and therefore record only 13CPchemical shifts in these experiments through the judicious choice of tf(-7 ms). Note that signals from glycine residues are still present, albeit of opposite sign relative to other cross peaks. A comparison of HNCACB spectra recorded with zf values of 3.5 ms (a) and 6.8 ms (b) (same measuring time for both spectra) is presented in Figure 7, where planes at an 15N
J. Am. Chem. SOC., Vol. 116, No. 26, 1994 11661
Backbone Assignment of lSN, I3C, and 2H Labeled Proteins
HN(CA)CB
CT-HN(C0)CA
I
i
-600
RZI
F72
V23
D24
US
L26
K27 N U
A29
YM Q31 N32 D33 L34
H3
Rerldue Figure 4. Strip plot of the CT-HN(C0)CA spectrum from R21 to H35. Only correlations connecting the l5N and HN chemical shifts with the I3Ca shift of the preceding residue are obtained from this data set.
tl b,
0
T r
.
@
Q
D33
: Q ,
4-t
A
I,
Q
b
'.?.
4
*
1.1
7.8
7.1
HN PPm
i')
'
I 8.0
* b
ta
an
- 60.0
8
. 12
7.0
8.0
. 1.8
l.43
. 7.8
.
.
.
7,1
1.1
7.0
L
j-
HN PPm
Figure 7. Comparison of HNCACB spectra recorded with zfvalues of 3.5 ms (a) and 6.8 ms (b). Equal measuring times were used for both spectra. A slice from each spectrum is presented at an 15N chemical shift of 116 ppm. Contour levels are spaced a factor of 2l'* apart in intensity. Only eight time points were recorded in the 15N dimension for each of the 3D experiments in this case.
I
J
,
8
0 70.0
chemical shift of 116 ppm from each experiment are shown. Recall that a tf value of -7 ms ensures virtually complete transfer from 13Cato 13CB carbons in the scheme of Figure 2c
so that only (13CP, 15N, HN) correlations are obtained while tf = 3.5 ms provides correlations to both 13Caand l3CP spins. An approximate factor of 2 increase in sensitivity of (13CP, 15N, HN) cross peaks (contours are spaced a factor of 21n in intensity apart) in the spectrum recorded with the longer tfvalue is noted. For this reason and because of the excellent sensitivity and spectral resolution provided by the CT-HNCNCT-HN(C0)CA schemes (Figures 3 and 4), separate experiments have been recorded to obtain 13Ca and 13CP correlations. Although the molecular mass of the trpR complex is 37 kDa, the protein portion of the complex is made up of a symmetric dimer with each monomer consisting of only 108 amino acids. Despite the predominantly a-helical nature of the molecule, the resolution afforded by the 3D methods described above is sufficient for the assignment of the I3Ca, 13CP, 15N, and HN chemical shifts. However, for applications to single chain large
Yamazaki et al.
11662 J. Am. Chem. SOC., Vol. 116, No. 26, 1994 I I I I I
7.0 .'
HN = 9.00 ppm N I 126.9 ppm
t
40
I
I
0.0
''O
I I I I
d -
I
HN = 8.44 ppm N 119.4 ppm
Cg = 37.8 ppm
QY \ 30 \
8.0.
I
Y30 0.0.
\
\
,.40
\
60
\
\
\
.60
.&
I I
:
HN
p
*+A29
I
C a I 61.0 ppm
so
CP
t :
.PPm
30
Cg-27.8ppm 0
8.0
1i
I
so HN = 7.32ppm N = 117.1 ppm
0.0 :
60
I '1.0
i :
8.0.
0.0.
= 52.5 ppm Cg-40.0ppm
~a
p""' I I I I I I
N
-
116.0 ppm 40
I
lso
Figure 8. Selectedplanes from the 4D HNCACB experiment at (I3Ca, I3CB)and (15N, HN)chemical shifts of residues A29-N32. The corresponding chemical shift pairs are indicated in the upper comer of each 2D data set. Note that A29 (*) is folded in the "C@dimension and the actual 13C@ chemical shift is 16.8 ppm. Although not indicated in the figure, all peaks folded in this dimension are of opposite sign relative to correlations that are not folded since a first-order phase correction of 180" is employed in the 13C@ dimension.
proteins, it may be advantageous to increase the dimensionality of experiments from three to four so that linkages of I3Ca, 13CP, 15N,and HN chemical shifts can be provided from a single cross peak in the 4D spectrum. It is strightforward to convert the 3D HN(C0CA)CB and HN(CA)CB experiments illustrated in Figures 2b,c into 4D experiments which record 13Ca,13CB, 15N, and HN shifts. In this case the two zfperiods immediately proceding the tl evolution periods in the 3D experiments are converted into cpnstant time evolution periods for recording 13Ca magnetization. Figure 2d illustrates the pulse scheme used for the 4D HNCACB experiment. Figure 8 shows a number of planes from the 4D HNCACB experiment at particular 13Caand 13CB chemical shifts as well as the corresponding 15N and HN planes. Each of the planes displays both intra- and inter-residue correlations providing a very straightforward approach for sequential assignment. Of course, in the absence of any degeneracies in (15N, HN) and (13Ca, 13CB) pairs of chemical shifts and assuming a complete data set, the assignment is, in principle, trivial. Often degeneracies can be sorted out by making use of the fact that particular amino acids have distinctive carbon chemical shifts.37 In order to quantitate the sensitivity of each of the experi-
ments, the S/N of each cross peak has been measured, and histograms showing the percentage of residues having a defined value of S/N are plotted in Figure 9. An estimate of the signal intensity of each peak was obtained from the peak height (measured using the software PIPP19) and the standard deviation in the noise estimated from a region of the spectrum without signal. Almost complete intraresiduecorrelationswere obtained in each of the CT-HNCA, HN(CA)CB, and 4D HNCACB experiments with over 95% of the expected interresidue correlations available from each of these experiments as well. In addition, nearly complete inter-residue correlations were observed in the CT-HN(C0)CA and HN(C0CA)CB experiments. The assignments of HN, 15N, 13Ca,and 13CPchemical shifts provided by the experiments described above are presented in Table 1. The assignments are consistent for the most part with those reported earlier for the trpWDNA complex containing the natural corepressor ~-tryptophan~* (5-methyltryptophan is the corepressor in the present study). A number of differences, (37) Grzesiek, S.; Bax, A. J. Biomol. NMR 1993,3, 185. (38)Zhang, H.;Zhao, D.; Revington, M.; Lee, W.; Jia, X.; Arrowsmith, C. H.;Jardetzky, 0.J . Mol. Biol. 1994,238, 592.
001
08
N/S ob
OZ
09
00
z P 9 9 01
ZI PI
001 Y
08
09
OP
oz u o o
.s
001
.01 .Sl
.
(e oz
oz
08 U ’
. .
-u-
oo
00L
.s .OL
. .
.si
.
02
08
11664 J. Am. Chem. SOC., Vol. 116, No. 26, 1994 Table 1. HN, 15N, and 13CB Chemical Shifts for Trp-Repressor Complex at 37 "C residue NH A2" E3 E4 S5
p6 Y7 S8 A9 A10 M11 A12 E13 414 R15 H16 417 E18 W19 L20 R21 F22 V23 D24 L25 L26 K27 N28 A29 Y30 431 N32 D33 L34 H35 L36 p37 L38 L39 N40 L41 M42 L43 T44 p45 D46 E47 R48 E49 A50 L51 G52 T53 R54 V55
8.58 8.48 8.44 ---8.00 7.90 8.52 8.10 7.90 8.09 8.15 8.20 8.40 8.20 8.03 7.80 7.90 8.02 7.84 7.92 8.16 8.34 7.31 7.56 8.99 8.42 7.30 7.32 7.89 8.22 9.06 7.47 ---7.95 8.91 8.16 7.67 8.57 7.67 9.80
----
8.16 8.63 7.66 7.61 8.10 8.00 8.09 7.90 8.20 8.11
I5N
13Ca
13CB residue MI
52.0 18.5 55.3 28.6 55.0 29.0 55.8 63.0 63.0 31.0 57.5 38.0 58.0 63.5 53.8 17.5 53.5 17.7 56.2 31.5 54.3 17.7 58.1 28.3 59.0 29.0 58.1 28.0 120.50 58.5 27.0 122.75 58.8 28.0 (57.7)b (28.8)b 123.13 58.0 28.4 117.50 56.9 39.2 120.88 58.9 28.5 121.25 61.0 36.6 119.75 65.4 30.2 121.63 56.5 40.8 124.29 57.2 41.2 119.88 56.5 39.0 120.83 59.2 31.5 116.78 55.2 38.0 126.92 55.1 16.8 119.38 61.1 37.2 117.15 57.2 27.5 115.73 52.5 40.3 117.54 55.8 39.5 118.36 53.0 42.0 121.15 58.6 28.0 119.43 59.1 36.4 - - - - - 64.7 30.0 120.01 58.0 41.0 120.90 57.8 40.5 118.60 54.1 37.2 120.09 57.2 41.2 115.30 55.1 31.0 115.70 53.2 44.1 117.88 58.6 67.1 - - - - - 65.1 30.8 116.75 56.7 38.5 118.63 58.2 29.0 117.13 59.8 29.0 117.01 58.8 28.8 122.29 54.5 16.5 119.05 57.8 40.1 106.30 46.4 - - - 120.13 66.2 68.0 119.75 60.0 29.4 118.63 65.5 30.3 122.38 122.38 119.19 ----119.75 117.88 126.50 121.25 118.81 123.31 119.38 119.75
R56 I57 V58 E59 E60 L61 L62 R63 G64 E65 M66 S67 468 R69 E70 L71 K72 N73 E74 L75 G76 A77 G78 I79 A80 T81 I82 T83 R84 G85 S86 N87 S88 L89 K90 A91 A92 P93 V94 E95 L96 R97 498 W99 Ll00 El01 E102 V103 L104 L105 K106 S107 D108
7.44 8.22 8.03 8.48 8.42 9.07 8.03 8.31 7.51 7.89 8.58 8.85 9.37 8.41 7.63 8.43 8.28 7.64 8.20 7.78 7.96 7.41 9.56 9.59 9.22 7.69 7.20 8.82 8.18 7.91 8.75 9.03 8.05 8.54 8.02 7.01 7.43 ---8.85 9.71 7.23 7.91 8.71 7.42 8.83 7.96 7.40 7.91 8.27 7.03 7.17 8.22 7.96
15N 13Ca 13C6 122.00 57.5 28.0 119.00 66.0 36.8 118.38 67.1 30.7 117.50 59.5 29.0 113.93 56.6 28.2 124.44 57.2 40.0 118.42 56.2 41.0 119.56 58.6 29.6 102.12 46.2 - - - 119.19 56.7 31.0 119.07 54.6 35.0 118.86 56.8 64.0 120.85 60.5 25.7 117.69 59.0 29.0 120.31 58.5 28.7 122.75 57.1 41.0 119.56 59.2 29.6 118.62 55.1 37.8 121.44 58.1 28.4 113.94 54.1 42.0 110.94 46.2 - - - 118.81 49.2 21.2 106.44 43.8 - - - 129.12 61.5 37.3 122.75 53.6 18.4 119.38 64.8 68.0 122.56 63.8 34.5 119.38 65.8 68.0 122.94 59.2 28.2 110.83 48.2 - - - 117.83 59.8 62.5 117.47 55.2 37.0 118.63 62.4 62.1 124.60 57.3 41.0 118.45 58.9 31.7 118.17 51.0 18.2 123.88 49.8 17.0 - - - - - 61.9 31.2 125.00 65.2 30.8 119.19 59.2 27.6 117.69 56.5 41.0 119.75 60.2 29.1 116.1958.227.8 121.44 61.2 28.0 117.88 57.0 41.6 116.94 59.1 28.5 118.44 58.2 27.8 113.75 63.6 31.0 117.88 55.5 41.5 115.62 54.2 40.0 119.94 55.8 32.0 118.06 57.5 63.8 127.81 55.1 41.4
-90% of the amino-terminal methionine of the original trpR translation product is removed.54 Tentative assignment.
of the form 4CxaC:N,, where Cxa is the transverse component of 13Camagnetization and C,' and N, are the z components of carbonyl and nitrogen magnetization, respectively. To a good first approximation the measured relaxation rate of 4C,cLc,'Nz is given by l/Tlemeasd= l/Tl,C.
+ l/TIC + l/TIN
where Tlemeasdis the value measured in the experiment, TleCa is the actual Tle value of the 13Ca spin of residue ( i - l), TlC is the selective T I value of the carbonyl spin of residue ( i - l), and TINis the 15N TI value of residue i. Therefore, in order to extract 13Ca Tle values from this experiment, it is necessary to measure 13C' and 15N TI values as well. In a previous we have measured the selective carbonyl UT1 rate in alanine
Yamazaki et al. dissolved in glycerol at 10 "C (correlation time of 16.7 ns) and obtained a value of 0.55 s-l. The dominant contribution to the selective UT1 rate is from 13Ca-13C' spin flips which scale as the overall correlation time. For the case of the trpR complex an overall correlation time of 14.5 f 0.9 ns is estimated on the basis of 15N T1/T2 ratios obtained from 36 residues that reside in well-structured regions of the molecule. Therefore an approximate value of 0.5 s-l ((14316.7) x 0.55 s-l) is estimated for l/TIC in the case of trpR. In contrast, the average value of l/TIN was measured to be 1.2 s-l for trpR. Since l/T1eca > l/TIN > 1/TlC' (see Figure l), we have elected to measure TINon a per residue basis and assume a uniform value of 0.5 s-l for 1/TlC'in the calculation of l/T1ec" from 1/TlQmd. The TleCavalues are converted to T2 values via the relation40 1
-
+
/ = sin2 ~ e ~ ( 1~ ~ ~cos2 ) e (I/T~)
COS'
e (I/T~)
(2)
where 8 = tan-l(Aoi/(yB1)}, Ami is the offset of 13Ca spin i from the carrier, and yB1 is the strength of the spin lock field. Equation 2 assumes that J(yB1) J(O), where J(w) is the spectral density function describing the motion evaluated at o. Representative Tle decay curves for both deuterated and protonated samples are indicated in Figure 1la,b. Figure 1IC compares 13Ca T2 values of the fully protonated trpR sample with T2 values of the 70%. deuterated sample where the contributions from the population of molecules with 13Ca spins one-bond coupled to 'H are eliminated. An average value for T2 of 130 f 12 ms is obtained for the deuterated sample versus 16.5 & 3.0 ms for the fully protonated sample. A T2 value of 16.5 ms agrees well with the value predicted for a molecule undergoing isotropic motion with a correlation time of 14.5 ns. (see Figure 1). However, the measured average 13Ca T2 value of 130 ms for 13Ca-2H pairs in the trpR complex is considerably lower than would be predicted assuming that the relaxation of the carbon spin is due exclusively to dipolar interactions with the attached deuteron. In this case a T2 value of approximately 250 ms is predicted. If one includes contributions to the relaxation of the 13Ca spin from the adjacent carbonyl and 13CP and 15N spins as well as CSA contributions (assuming an anisotropy of A o = 34 ppm41), a 13Ca T2 value of -160 ms is calculated. In addition, contributionsto relaxation can also arise from neighboring (nonbonded) IH dipolar interactions. For example, relaxation from a single 'HP proton would decrease the 13Ca Tz value to -135 ms. Although deuteration of proteins as a vehicle for increasing the molecular mass limitations of current NMR methods has obvious benefits, it is not without problems. Aside from the obvious drawback associated with recording 'H- 'H NOE spectra on molecules with a depleted set of 'H spins, the substitution of deuterons for protons results in an increase in
-
(40) Peng, J. W.; Wagner, G. J . Maan. Reson. 1992,98, 308. (41) Naik A.; Ganaphy, S.; Akaska, K.; McDowell, C. A. J . Chem. Phys. 1981,74, 3190. (42) H e m , E. R.; Szabo, A. J . Chem. Phvs. 1985,82, 4753. (43) Moms, G. A.; Freeman, R. J . Am. Chem. SOC.1979,101, 760. (44) McCoy, M.; Mueller, L. J. Am. Chem. SOC. 1992,114, 2108. (45) Shaka, A. I.; Keeler, J.; Frenkiel, T.; Freeman, R. J . Magn. Reson. 1983,52,335. (46) Kay, L. E. J. Am. Chem. SOC. 1993,115, 2055. (47) Boyd, J.; Scoffe, N. J . Magn. Reson. 1989,85, 406. (48) Patt, S . L. J . Magn. Reson. 1992,96, 94. (49) Geen, H.; Freeman, R. J . Magn. Reson. 1991,93, 93. (50)McCoy, M.; Mueller, L. J. Magn. Reson. 1992,98, 674. (51) Marion, D.; h a , M.; Tschudin, R.; Bax, A. J. Magn. Reson. 1989, 85, 393. (52) Kay, L. E.; Nicholson, L. K.; Delaglio, F.; Bax, A.; Torchia, D. A. J . Magn. Reson. 1992,75, 699. (53) Palmer, A. G.; Skelton, N. J.; Chazin, W. J.; Wright, P. E.; Rance, M. Mol. Phys. 1992,97, 359. (54) Joachimiak,A.; Kelley, R. L.; Gunsalus, R. P.; Yanofsky, C.; Sigler, P. B. Proc. Natl. Acad. Sci. U S A . 1983,80,668.
Backbone Assignment of 15N, 13C, and zH Labeled Proteins I
L \
I
J. Am. Chem. SOC., Vol. 116, No. 26, 1994 11665 v
-v
"
m,
Figure 12. Pulse scheme for the measurement of HN TI values. AU
u 0.18
c r f :
0.16-
x
-t
2
i
0.1
c" 0.08
o.041
0.061
0.02
so
-30
40
i
50
60
io
60
90
100
110
Residue Figure 11. Comparison of I3Catransverse relaxation times for 13Ca2H pairs in the 15N, 13C,2H trpR complex relative to 13Ca-lH pairs in the 15N, 13C trpR complex. (a) Fits of Tle data measured with the sequence of Figure 10 for 13Ca-2H pairs of residues E65 (0),A91 (x), and L105 (*). (b) Corresponding fits of Tle data for residues in the fully protonated sample. In order to minimize overlap of the decay curves, 0.1 has been added to the intensities of the peaks from A91 while 0.1 has been subtracted from the intensities of peaks associated with E65. (c) Comparison of T2 values as a function of residue in deuterated(*) and protonated (0)Q R . T2 values were calcualted from T I , measurements using eqs 1 and 2, with the simplification that UTI, cos2 0 (UT& Average T2 values of 130 and 16.5 ms are obtained from deuterated and protonated samples, respectively.
-
the T I values of the remaining protons. Since the great majority of NMR experiments developed for protein structure determination, including all of the methods described here, begin with magnetization originating on protons, the lengthening of proton T1 values can have a significant deleterious effect on the sensitivity of these experiments. The severity of the effect will clearly be dependent on the protein studied as well as the level of deuterium incorporation. For experiments where the magnetization begins on the HN proton (all of the experiments considered in this paper), saturatioddephasing of water magnetization should be particularly avoided so that the water is always in a near-equilibrium state. In this way the water can serve as a rich supply of magnetization through exchange with the labile HN protons, thereby increasing the effective rate at which HN magnetization is restored to eq~ilibrium.~O-~~ In order to investigate the increase in relaxation times associated with 70% deuteration in trpR, we have measured HN TI values for 15N, 13C, 70% 2H trpR and compared these values with TI values measured for a protonated 15N, 13C sample. Figure 12 illustrates the pulse sequence that was used for the HN TI measurements. The 90, (water selective) spin lock 90-, (water selective) portion of the sequence effectively dephases all of the protein magnetization while restoring the
narrow (wide) pulses have a flip angle of 90" (180"). Unless otherwise indicated the phase of all of the pulses is x. The shaped 'Hpulses are either 2 ms 90" rectangular pulses or 2 ms pulses having the SEDUCElU profile and are selective for water. A 7 kHz C.W. pulse (along the x axis) is applied for 60 ms between the first two shaped pulses to completely randomize 'Hmagnetization from the protein so that at the start of the recovery period, T, the protein magnetization has the initial condition M L= 0, while for water the magnetization is close to the equilibrium value. 15Ndecoupling during acquisitionis achieved using a 1 kHz WALTZ45field, while 15Npulses are applied using a 5.3 kHz field. The carbon 180' pulse is applied using an 18 kHz field centered at 117 ppm, approximately midway between the 13Caand the 13C' resonances. The delays used are z, = 2.4 ms, 6 = 1.4ms, 6' = 0.5 ms. The phase cycling employed is 41 = x; & = x,-x; $3 = 2(~),2(-~),2(-~),2(-y); 4 4 = x ; rec = 2(x),2(-x). For each value of rI, N- and P-type coherences are obtained by recording two data sets whereby the sign of the gradient g4 is inverted and 180" added to the phase of 44 for the second data set. Data sets obtained for positive and negative g4 values are stored in separate memory locations and pure absorptive line shapes in the 15Ndimension generated by adding and subtracting the N- and P-type data sets, storing the new data sets separately, and applying a 90" zero-order phase correction in the acquisition dimension to one of the new data sets.25 The phase of $1 and the phase of the receiver are incremented by 180" for each increment of t1.51 The durations and strengths of the gradients (rectangular) are g1 = (1.0 ms, 8 G/cm), g2 = (0.5 ms, 8 Gkm), g3 = (1.5 ms, 15 G/cm), g4 = (1.25 ms, 30 G/cm), g4' = (0.125 ms, 27.8 G/cm), g5 = g6 = (0.4 ms, 4 Gkm). water to the +z axis immediately prior to a delay T, during which time the recovery of protein magnetization occurs. Thus the initial state of the protein magnetization is M, = 0, while the water magnetization is near equilibrium and resides along the +z axis. This is similar to the magnetization state that exists in the triple resonance experiments of Figure 2 at the start of the acquisition period. The remaining portion of the sequence is the enhanced sensitivity HSQC pulse scheme with gradients employed to select for N-type or P-type transfer pathways, described p r e v i ~ u s l y .Note ~ ~ that saturatioddephasing of water magnetization is minimized throughout the sequence. Although the relaxation of protons in proteins is a multiexponential process due to cross relaxation with other protons, the recovery curves are nevertheless fit with a single exponential, as Figure 13a,b indicates. A comparison of the HN TI values measured for 15N, 13C, 70% 2H and 15N, 13C trpR samples is provided in Figure 13c. On average the HN TI values increase by a factor of 1.85 from 0.7 f 0.2 s to 1.3 f 0.3 s on deuteration to a level of -70%. The results presented clearly indicate that spectra of high sensitivity and resolution can be obtained with 15N, 13C, 2H samples of proteins having molecular masses on the order of 40 kDa. It is also important to establish that these experiments cannot be performed successfully on fully protonated samples. Figure 14 illustrates comparative one-dimensional spectra of fully protonated (a) and -70% deuterated (b) 15N, 13C labeled trpR recorded using the pulse scheme of Figure 2c, HN(CA)CB normalized as described in the legend to the figure. In the absence of relaxation effects, the spectra should be of similar intensity. Clearly, this is not the case, and for the most part, only signals from residues of the mobile regions of trpR show
Yamazaki et al.
11666 J. Am. Chem. SOC.,Vol. 116, No. 26, 1994 8
I
I
7
6
5
5
.e
$ 4 3
3
2
.
.
10.5
.
.
.
.
10.0
.
.
.
9.5
.
.
.
.
9.0
.
.
.
.
8.S
.
.
.
.
8.0
.
.
.
.
7.5
.
.
.
.
7.0
.
.
.
.
.
.
.
.
.
6.5
1
'
0
2
4
6
8
Time (s)
Time (s)
/ ' " ' 1 ' " ' 1 ' ' ' ' 1 ' ~ ' " ' " 1 " " 1 ' " ' ~ ' " ' 1 ' ' " ' " ' '
10.5
0.21 $0
30
a0
50
60
io
80
sb
& 1M)
I 1;o
Residue Figure 13. Comparison of HN TI values of (a) the 15N, 13C, -70% *H trpR complex and (b) the 15N,13CtrpR complex. Single exponential fits of the recovery data for residues D24 (0),N87 (x), and K106 (*) are shown. (c) HN TI values as a function of residue for the deuterated (*) and protonated (0)Trp complexes. Average T I values of 1.3 and 0.7 s are obtained for deuterated and protonated samples, respectively.
significant intensity in the spectrum of the fully protonated sample. It should be noted that equal relaxation delays (2 s) were employed in the recording of both spectra and that the difference in HN TI values of a factor of -1.8 suggests that spectra recorded on the fully protonated sample can be measured using approximately one-half the relaxation delay (1 s) employed in the recording of spectra of the deuterated sample (2 s). Using average NH T I values of 1.3 and 0.7 s for the deuterated and protonated samples, respectively, an increase in the relative sensitivity of the protonated spectrum of -15% can be realized in this case. Nevertheless, the difference in the quality of the spectra is significant, illustrating the importance of deuteration. Because in all of the experiments presented in this paper the time that transverse magnetization resides on the 13Ca spin is approximately the same (-26-27 ms), similar comparative spectra are obtained with the other pulse schemes as well.
Concluding Remarks In this paper a number of triple resonance pulse schemes for the assignment of backbone 15N, HN, 13Ca as well as l3C8
10.0
9.5
9.0
8.5
8.0
7.5
7.0
6.5
HN PPm Figure 14. Comparative 1D spectra of fully protonated (a) and -70% deuterated (b) I5N, I3ChpR using the HN(CA)CB scheme (Figure 2c). A total of 1024 transients were recorded for the 2H sample and 6928 transients (1024 x 2.62) for the 'H sample which is a factor of 2.6 more dilute. The rms noise levels are normalized in the plots. chemical shifts have been presented. The methods have been demonstrated on a ternary complex of 15N,13C,-70% 2H trpR, 5-methyltryptophan, and a 20 basepair trp-operator DNA fragment (MW 37 000). The use of partial deuteration in concert with deuterium decoupling allows exploitation of the long 13CaTz values, for the case of 13Ca-2H pairs, so that highsensitivity spectra can be obtained. In addition, for CT-HNCA and CT-HN(C0)CA spectra, the use of constant time carbon evolution provides for significant increases in resolution in this dimension relative to previous versions of the experiments. These experiments should provide a useful approach for backbone assignment of many proteins in the 30-40 kDa molecular mass range where experiments such as the HNCACB on fully protonated samples often fail or provide only partial connectivities. Current efforts are directed toward the development of experiments for extending the assignment of all of the side chain 'H and 13C spins.
Acknowledgment. This research was supported by the National Cancer Institute of Canada (C.H.A and L.E.K), by the National Sciences and Engineering Research Council of Canada (L.E.K), and by the Human Frontiers Science Program (C.H.A). Toshio Yamazaki is the recipient of a Human Frontiers Science Program Fellowship.
